[jira] [Comment Edited] (IGNITE-9784) Performance drop on sql yardstick benchmarks
[ https://issues.apache.org/jira/browse/IGNITE-9784?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16639476#comment-16639476 ] Alexey Goncharuk edited comment on IGNITE-9784 at 10/19/18 9:46 AM: The regression is caused by a more frequent checkpoint triggering caused by IGNITE-6552. Re-check with increased WAL history size confirms that there is no performance drop. As a solution, we should choose default history size as 4x of checkpoint buffer size (unless overridden by a user). was (Author: agoncharuk): The regression is caused by a more frequent checkpoint triggering caused by IGNITE-6552. Re-check with increased WAL history size confirms that there is no performance drop. As a solution, we should choose default history size as 3x of checkpoint buffer size (unless overridden by a user). > Performance drop on sql yardstick benchmarks > > > Key: IGNITE-9784 > URL: https://issues.apache.org/jira/browse/IGNITE-9784 > Project: Ignite > Issue Type: Bug > Components: persistence >Affects Versions: 2.7 >Reporter: Ilya Suntsov >Assignee: Alexey Goncharuk >Priority: Blocker > Fix For: 2.7 > > Attachments: Screen Shot 2018-10-03 at 4.52.22 PM.png, > background_primary_sync.png, backgroung_full_sync.png, fsync_full_sync.png, > fsync_primary_sync.png, log_only_full_sync.png, log_only_primary_sync.png > > > Please take a look at the attachment. > Grid configuration: > * 4 servers / 8 clients > * java 8 > * 64 threads, 60s warmup, 180s duration -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (IGNITE-9784) Performance drop on sql yardstick benchmarks
[ https://issues.apache.org/jira/browse/IGNITE-9784?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16639476#comment-16639476 ] Alexey Goncharuk edited comment on IGNITE-9784 at 10/9/18 8:40 AM: --- The regression is caused by a more frequent checkpoint triggering caused by IGNITE-6552. Re-check with increased WAL history size confirms that there is no performance drop. As a solution, we should choose default history size as 3x of checkpoint buffer size (unless overridden by a user). was (Author: agoncharuk): The regression is caused by a more frequent checkpoint triggering caused by IGNITE-6552. Re-check with increased WAL history size confirms that there is no performance drop. As a solution, we should choose default history size as 2x of checkpoint buffer size (unless overridden by a user). > Performance drop on sql yardstick benchmarks > > > Key: IGNITE-9784 > URL: https://issues.apache.org/jira/browse/IGNITE-9784 > Project: Ignite > Issue Type: Bug > Components: persistence >Affects Versions: 2.7 >Reporter: Ilya Suntsov >Assignee: Alexey Goncharuk >Priority: Blocker > Fix For: 2.7 > > Attachments: Screen Shot 2018-10-03 at 4.52.22 PM.png > > > Please take a look at the attachment. > Grid configuration: > * 4 servers / 8 clients > * java 8 > * 64 threads, 60s warmup, 180s duration -- This message was sent by Atlassian JIRA (v7.6.3#76005)